Multi-modal imaging: simultaneous EEG-fMRI

1. Silvina G Horovitz, PhD Human Motor Control Section Medical Neurology Branch National Institute of Neurological Disorders and Stroke National Institutes of Health Multi-modal imaging: simultaneous EEG-fMRI

2. Outline • EEG overview • Why simultaneous EEG-fMRI? • How? Technical considerations • When? Examples

3. EEG (electroencephalography) measure of synchronous activity of population of neurons,primarily reflects postsynaptic potentials.

4. EEG measures Isolated amplifiers filters A/D converter recording device Eyes closed Electrodes and conductive media 1 s Acquisition PC

5. montage International 10-20 System of Electrode Placement F - Frontal lobe T - Temporal lobe C - Central lobe P - Parietal lobe O - Occipital lobe "Z" refers to an electrode placed on the mid-line. Odd: left Even: right Electrode configuration • Referential • (Si vs. Ref; Sk vs. Ref) • Bipolar • (Si vs.. Sk )

6. Data processing • Time domain • Event Related Potentials (ERPs) • pre-processing: • detrend - filtering • epoch • baseline correction • ocular artifact reduction • (common grounded, artifact rejection) • time-locked averaging

7. Data processing • Frequency domain • Power at different bands • Power spectra density (FFT) • Cross-spectra (correlation among different electrodes) • Coherence (measure of stability of the phase shift between electrodes) • Event related desynchronization

8. Human EEG Spontaneous Evoked Depth ECoG Transient Steady State Scalp Clinical Applications epilepsy, head trauma, drug overdose, brain infection, sleep disorder, coma, stroke, Alzheimer’s disease, brain tumor, multiple sclerosis, surgical monitoring Cognitive Science sensory pathways, stimulus encoding, motor process, spatial task, verbal task, mathematics, short term memory, memory encoding, selective attention, task context, general intelligence, dynamic brain theory PL Nunez, EEG, Encyclopedia of the Brain, 2003

9. Why do we want to measure EEG and fMRI simultaneously?

10. Neuroimaging BRAIN ACTIVITY f(x,t) fMRI k{g[f(x,t)]} Metabolic/ vascular responses g[f(x,t)] • Poor time resolution (s) Good spatial resolution (mm) electrical EEG, MEG m[f(x,t)] Good time resolution (ms) Poor spatial resolution

11. P2 P1 N170 EEG fMRI EEG is the gold standard for sleep studies, epilepsy, some cognitive tasks, etc

12. HOW often measured together? Data PubMed search July 14, 2015 Nr. of publications YEAR

13. EEG setup EEGstudy Stimulus PC Acquisition PC EEG amplifier

14. EEG-fMRI setup Amplifier EEG-fMRIstudy Optic fiber Acquisition PC Adapter to projector Synchronization Scanner clock Stimulus PC Scanner trigger

15. Technical Issues Electromagnetism 101 • Maxwell's Laws. • A changing magnetic field produces an electric field • A changing electric field or current produces a magnetic field. BIG PROBLEM Luckily, the magnetic field change Form the EEG does not affect the image quality!

16. THE not so good NEWS • MRI is noisy • Electrical noise  MRI and EEG were not meant for each other … Remember Maxwell's Law?

17. Simultaneous EEG-fMRI - Technical issues Example from BOLD & Perfusion MRI sequence optimized for EEG-fMRI acquisition 5 slices • Sources of artifacts: • gradient artifact • physiological noise: ballistocardiogram

18. Simultaneous EEG-fMRI - Technical issues Approximate values of different signals • Gradient artifact : ± 10mV • EEG: ± 150mV • BC artifact: ± 200mV • Blink: ± 150mV • Movement: < 1mV • ECG: ± 20mV • EMG: ± 50mV • Helium pump: 40-60Hz and

19. THE good NEWS • MRI compatible EEGequipment, leads and electrodes • Safe for the scanner • Safe for the subject More on safety later • Careful setup: • Equipment • Cables • Subject head

20. DATA acquisition • Sample EEG at 5 kHz (or more) • Slice TR at a frequency that is not of interest (and a round number) • Low Pass Filter at 250Hz • ~0.01 Hz high pass to avoid saturation (use DC only if enough range ) • Volume (or slice) marker • Resolution: 0.5mV (make sure dynamic range covers the signal, depends on scanner and configuration) • Clock synchronization

21. EEG DATA acquisition • Make sure amplifiers do not saturate • Adjust amplifier resolution • Keep electrodes’ impedance low (unless using high impedance equipment) • Keep cabling safe and fixed • Have a good cardiac signal • Adjust MR sequence • Adjust experiment (ISI <> TR)

22. Gradient artifact removal • Artifact removal (Matching filter) • Create a template of the artifact • Subtract average artifact • If proper timing • and no movement works great! Raw EEG during EPI acquisition

23. Raw Gradientartifact corrected Dr Jen Evans

24. Ballistocardiogram artifact removal Matching filter (BV Analyzer) (Allen et al, 2000): Detect R Create a template Subtract (allows for amplitude adjustment) Single Value Decomposition (Neuroscan) Run classification Remove components Reconstruct time series Optimal base set (EEGLAB Niazy, 2005) PCA to create bases Fitting (adaptive algorithm) Subtraction Combinations i.e  Liu, 2012 use ICA, SVD & mutual information (based on Peng, IEEE 2005) software download: http://amri.ninds.nih.gov/cgi-bin/software

25. Gradient artifactcorrected Cardioballistogramcorrected Dr Jen Evans

26. Gradient artifactcorrected Cardioballistogramcorrected

27. SAFETY considerations Simultaneous Electroencephalography-Functional MRI at 3 T: An Analysis of Safety Risks Imposed by Performing Anatomical Reference Scans With the EEG Equipment in Place Ulrike Nöth, Laufs, Stoermer, and Deichmann JMRI 2012 SAR: Specific Absorption Rate (or the energy deposited in the body by the radio frequency transmission)

28. SAFETY considerations • Sequences • EPIs (in most cases ok to run an MPRAGE for localization) • be aware of high res short TR EPIs (pay attention to SAR) • Special sequences require special safety testing • Set up • Cables straight and in the center. • Avoid loops • Equipment as far back from iso-center as possible • (far front for EMG) • All scanners are not equal; gradients and coils affect electrodes’ temperature • Be aware different body shapes and weight load coil differently

29. Interim Summary • EEG measurements have: • Good temporal resolution • Poor spatial resolution (when measure non invasively) • Electrical and hemodynamic responses are related • Simultaneous EEG-fMRI requires special equipment • SAFETY PROCEDURES ARE KEY • Dimensionality reduction is needed for data integration

30. When do we want to measure EEG and fMRI simultaneously?

31. When is it important to measure simultaneously? • State dependent analysis • Alertness • State vs Trait • understanding of BOLD signal/EEG origings • Physiological markers defined by EEG • Seizures • Sleep stages

32. Type of studies • Correlations of EEG and fMRI • In time domain • In frequency domain • Multivariate methods • ICA • Informing one with the other • Sorting data and perform analysis in one modality • Mix analysis

33. EEG parameter as regressor

34. BOLD-EEG band-power correlations Goldman et al. 2002 Simultaneous EEG and fMRI of the alpha rhythm.

35. How to link time and space information? Parametric studies and correlational analysis Correlation maps between fMRI signal change and P300 amplitude. Composite of 7 subjects. Horovitz et al MRI, 2002 OLD DAYS: SAME SUBJECTS, EEG AND FMRI ON SEPARATE SESSONS

36. Assessing the spatiotemporal evolution of neuronal activation with single-trial event-related potentials and functional MRI Tom EichelePNAS 2005 vol. 102 no. 49

37. Single-Trial Analysis of Oddball Event-RelatedPotentials in Simultaneous EEG-fMRI Benar et al. Human Brain Mapping 28:602–613 (2007)

38. Correlation between Amplitude of BOLD fluctuations and alertness Index derived from EEG DAYTIME 1 hour PARADIGM Horovitz et al HBM, 2008

39. EEG to define states

40. Changes in the level of consciousness Use EEG to sort fMRI data

41. Do changes in connectivity over time have a physiological origin?

42. EEG-vigilance and BOLD effect during simultaneous EEG/fMRI measurement S. Olbrich et al. / NeuroImage 45 (2009) 319–

43. Spatiotemporal dynamics of the brain at rest-exploring EEG microstates as electrophysiological signatures of BOLD resting state networks. Yuan H, Zotev V, Phillips R, Drevets WC, Bodurka J. Neuroimage. 2012    

44. EEG to understand BOLD signal

45. Correlations of simultaneously acquired SSVEPs with BOLD fMRI response J.W Evans et al OHBM 2015

46. EEG & fMRI to study disease

47. Widespread epileptic networks in focalepilepsies—EEG-fMRI study Fahoum et al Epilepsia, 53(9), 2012 Group analysis results for hemodynamic response functions peaking at 3, 5, and 7 s after the interictal epileptic discharges Differentepileptic syndromes result in unique and widespread networks related to focal IEDs. TLE group (n = 32) FLE group (n = 14) Between April 2006 and December 2010, 168 consecutive 3T EEG-fMRI scans were performed in focal epilepsy patients

48. intracranial recordings - fMRI Simultaneous intracranial EEG–fMRI in humans: Protocol considerations and data quality D.W. Carmichael (2012)

49. Simultaneous EEG-fMRI summary • Safety first! • Quality control at experiment setup & data collection • Equipment setup • Pulse sequence • Task design • EEG pre-processing • Gradient & ballistocardiogram artifacts • Data integration • Dimensionality reduction • Spatial correlations • Regressions • Sorting data based on state • Some applications • Understanding BOLD signal • Understanding Disease • Origins of EEG signals • State dependent studies